‘Liver transplant 3D modeling platform cuts physicians’ burden, raises care quality’

Korean doctors have thrown down the gauntlet to the Japanese-dominated 3D modeling-based surgical planning software market.

They are four Samsung Medical Center physicians -- Professors Ryu Jin-soo and Oh Nam-kee of the Department of Transplantation Surgery and Professors Jeong Woo-kyoung and Kim Jae-hun of the Department of Radiology. They developed a “liver transplant AI model,” which automatically measures the size and capacity of a liver donor's liver based on CT images.

Previously, transplant surgeons had to segment the donor's liver according to its anatomy based on CT images and calculate the volume manually, which was very time-consuming for doctors. Even if 3D modeling-based surgical planning software was available, its high price and low ease of use limited its universal use.

Korea Medical Review spoke to Professor Ryu to learn about the process of developing the 3D modeling platform.

Question: From publishing comics to offering VR training on liver transplantation, and to developing an AI model for liver transplantation, you don’t seem to hesitate to try new things. Why is that?

Answer: I'm just the type of person who wants to create something new, especially something that hasn't been done before. When I feel like I want to do something, I just start thinking about it and go for it.

I started drawing comics because I've always been interested in drawing, and when I became a doctor and practiced medicine, I realized that it would be great to create content about these things. When I started to incorporate VR and AI into medical treatment and surgery, I drew the structure in my head while preparing for the surgery and used a 3D model, and I thought that it would be possible to reduce the work of humans by using VR and AI. I think it came naturally to me as I grafted it on one by one.

Q: Was there any transplant case that impressed you?

A: It's always meaningful to see patients recovering from liver transplants. One case that stands out in my mind is the recovery of a young patient who was in a really desperate condition. The patient was a North Korean defector. His liver was destroyed by chronic rejection after a liver transplant, so he was transplanted with a brain-dead person’s liver. Because it was a re-transplant, the surgery was very difficult, but he recovered well and is now very healthy. I also remember a patient who had surgery as a child for congenital biliary atresia, but his liver gradually failed, so he had a re-transplant as an adult. He had severe intestinal adhesions from the previous surgery and had to undergo several reoperations for intestinal perforation, and his liver failed again, so he had to undergo a re-transplant. Fortunately, he recovered well and is living a healthy life.

However, I remember many cases where the patient did not recover well. When I think back on those cases, I feel regret, but I try to think about what I could have done to change the outcome and use it as a reference for future treatments.

Q: What inspired you to develop an AI model of a liver transplant?

A: This AI model is an automatic 3D model of the liver. When performing liver surgery, 3D modeling is very helpful for precise surgical planning and safe operation, and we have been working on creating our 3D models for five years. It's a medically and technically complex task, not just creating a 3D model, but customizing the image processing according to the surgical plan. We decided to combine AI models with this because we thought it would reduce the amount of human work and produce more results faster, and if the AI model is successful, it could be used in large hospitals and small-and medium-sized hospitals.

Q: What were the challenges in developing the AI model for liver transplantation?

A: We started developing the AI model in 2020, but it was difficult to find people to work with because the methodology for training AI models was not universalized or optimized. We approached AI researchers, but were repeatedly rejected due to lack of data.

After the learning methodology was improved and after we formed a team with Professors Jeong Woo-Kyoung and Kim Jae-hun from our hospital's radiology department, we could develop an AI model for pancreatic biliary tract imaging. Still, it was difficult to collect a lot of cases and increase the data. The most important thing in developing a good AI model is good data. We didn't create 3D models for the purpose of developing AI models but we used 3D models from actual surgeries to develop AI. Since the number of high-quality data was limited, it took a long time to collect the data, and we had to develop different AI models for different types of images, not only CT but also MRI.

Q: How is the liver transplant AI model being used in actual surgery?

A: We use the AI model for automatic 3D modeling, and then re-process the data to use it as 3D modeling content for the surgery. When planning how to cut before surgery, the 3D model is used to understand where we are cutting during surgery, and the AI model is used to automatically convert the 3D model.

Previously, liver volume measurements of living liver donors were done manually by specialists, but this year, the AI model has been used for screening. That reduced the time and labor intensity, also lowering the error rate, which can vary slightly from doctor to doctor. Of course, doctors make the final decision.

Q: What is the biggest change in liver transplant surgery that will come from commercializing AI models for liver transplantation?

A: We started a startup called “LiverAIz” to provide patient-specific 3D modeling using AI models as a service. Our AI models and know-how in processing 3D models customized to the information needed by liver surgeons are being jointly developed with a healthcare startup called “SURGICAL MIND” as an AI-based 3D model surgeon planning platform.

Many existing 3D model software are made in Japan and end up requiring doctors to fill in the data manually one by one. We are planning a service that allows doctors to upload data and automatically create the desired 3D model to plan surgery and refer to it during surgery. Such a service can dramatically reduce the workload of doctors and improve their workflow.

Also, existing 3D model software is quite expensive, costing tens of millions of won. Once our service is commercialized, the cost burden will be significantly reduced, and 3D modeling-based surgery will become commonplace for liver transplants and liver resections, even in small- and medium-sized hospitals and rural hospitals. It has already been established in several papers that 3D modeling-based surgery is much safer than surgery based on conventional imaging. We believe that our service will enable safer and more accurate surgeries and provide patients with improved care.

Q: What are the biggest barriers to commercializing AI models for liver transplantation, and what are the solutions?

A: It is important to create a cost structure for 3D-based surgical procedures, which is a different way of doing surgery. In the U.S., we pay a lot of money to use these 3D modeling services to ensure patient safety. Japan has long established a fee for 3D modeling of liver surgery.

In Korea, 3D-based surgical procedures are not included in the reimbursement, so only a few large hospitals use existing 3D model programs. Rather than purchasing such a program, our goal from the beginning was to create our own 3D modeling platform by developing our AI model.

Above all, it is important for 3D modeling-based surgery to be recognized as a new medical technology in Korea. At this point, hospitals that require 3D modeling pay for it in liver transplant surgeries. The majority of liver cancer patients or liver transplant patients cannot benefit from 3D model-based surgical procedures if the hospital pays for it.

Q: What do you hope to accomplish with this AI model for liver transplantation?

A: The product that LiverAIz and SURGICAL MIND are jointly developing is targeted for release early next year. It aims to combine the know-how of Samsung Medical Center's 30 years of experience and five years of research on 3D modeling.

We want to support busy liver transplant surgeons as much as possible so they can focus on their patients. Most professors have a strong desire to perform surgeries using 3D modeling. However, even if the hospital supported 3D modeling surgery, they had to use Japanese products.

We want to make a product made in Korea and Korean doctors can use comfortably, so that doctors who want to use 3D modeling in surgery can feel comfortable and trust it. Ultimately, our goal is to create an environment where our product can be used anywhere in the country, not just in a few hospitals, so that patients can experience the same level of medical quality no matter where they are in the country.

And because what we're building is a platform, we want to expand our pipeline beyond the liver to other organs. 3D-based surgical planning is the same format for any organ. We have started discussions to utilize our platform for other organs such as gallbladder, pancreas and lungs.

Q: What are your goals for the future?

A: I want to focus on spreading the knowledge and technology that I have studied and accumulated and creating new value. I want to become a physician entrepreneur, which is a different type of doctor from those who operate and see patients in a hospital. I think that launching a startup will be an opportunity to spread good influence beyond the hospital where I work. My immediate goal is to treat patients, find good methods and make them available to others, and to do that. I want to take it as far as I can and go as far as I can with the identity of an entrepreneur.